Inaccuracy of ICD-9 Codes for Chronic Kidney Disease: A Study from Two Practice-based Research Networks (PBRNs)

Discussion

Based on our findings, nearly half of patients in the registries did not have the clinical indicators to support a diagnosis of CKD stage 3, suggesting that registries developed from ICD-9 codes may not accurately represent the prevalence of CKD stage 3 in these practices.

The reasons behind this high prevalence of misdiagnosis cannot be defined by this study. However, this finding raised many questions as to why this may be occurring. Some possible reasons include CKD-specific issues, alternative uses of ICD-9 codes, and challenges inherent in the practices where this study was conducted. These issues are discussed below.

Studies have suggested that primary care physicians are uncomfortable with or unaware of the KDOQI guidelines for CKD, especially with patients in the early to moderate stages of the disease.^7⇓–9 KDOQI guidelines state that a patient meets criteria for CKD stage 3 if they have had 2 estimated glomerular filtration rates <60 at least 90 days apart or have had 2 albumin creatinine ratios >30 at least 90 days apart.^10,11 Thus, accurate diagnosis relies on 2 abnormal lab tests observed in the appropriate time frame. However, our findings suggest that physicians may be diagnosing CKD after only 1 abnormal test result. Furthermore, many of the misdiagnosed patients had normal or insufficient test results, suggesting that physicians may be diagnosing based on factors not supported by the KDOQI guidelines¹² (see Table 4).

Another explanation for our finding involves the use of ICD-9 codes in ways other than to signify a diagnosis. ICD-9 codes may be used by practitioners to signify a “working diagnosis,” to allow a test or procedure to be covered by a patient's insurance or to increase reimbursement for the practice.^1,13 Billing staff may add or modify codes for the same reasons.

This high prevalence of CKD misdiagnosis that we observed might be related to the characteristics of the practices where we conducted chart reviews. All 3 practices are residency teaching sites in family medicine. Residents may be even less comfortable with CKD guidelines than attending physicians, leading to more inaccurate diagnoses.⁷ In addition, residency practices experience less continuity of care because of student turnover. Reduced continuity of care could contribute to the perpetuation of an inaccurate diagnosis in the EMR because physicians may be unlikely to question a diagnosis made by a previous provider. These practices also are located in low-income neighborhoods where physicians treat patients with complex medical and social problems. These competing demands could potentially distract physicians from guideline-concordant CKD diagnosis for patients in the early stages of the disease.

Although we cannot yet determine the reasons behind this issue, it has important implications for both research and patient care. Electronic databases built from registries of patients with certain diagnoses, as tracked by ICD-9 codes, are increasingly used for research.^4,13 If a large percentage of these ICD-9 codes, such as the ones found in this study, do not reflect the actual presence or prevalence of disease, the research findings will be skewed. Outcomes research may be tainted and irreproducible if the patients identified in these registries do not have the disease in question.⁴

In addition, if the prevalence of misdiagnosis we observed in this patient cohort is present in other populations, thousands of patients may be mismanaged. This creates an enormous patient safety problem, since patients may be receiving unnecessary tests and/or treatments. For example, a patient treated with an angiotensin-converting enzyme/angiotensin receptor blocker for an erroneous diagnosis of CKD may be unnecessarily exposed to the risk of hyperkalemia, a serious adverse event.

This study has several limitations. First, we reviewed only charts of patients who already had a diagnosis of CKD stage 3. Our study did not investigate patients who met criteria for CKD but did not have a corresponding ICD-9 code in their records. The under-recognition of CKD is a well-documented problem, which is why were surprised to find evidence of overdiagnosis among this population.⁸ The clinical data from a limited population (n = 325) makes it difficult to determine the reasons behind our findings. A strength of this study, however, was that we were able to find similar results among 3 separate practices in 2 different regions of the country. The practices we chose had similar patient populations, and all 3 were residency teaching clinics. We believe this is a strength of our study. Our findings suggest, however, that these results might be replicated in practices similar to those we studied. This would implicate thousands of misdiagnosed patients. More research needs to be done to determine whether these findings are reproducible in a broader range of primary care practices. If this is confirmed, practices similar to those we studied may be affected. Further research should be undertaken to uncover the reasons behind the high prevalence of misdiagnosis, and initiatives should be developed to address this issue.